Article Text

Download PDFPDF

The long-term prognosis of cardiovascular disease and all-cause mortality for metabolically healthy obesity: a systematic review and meta-analysis
  1. Ruizhi Zheng1,
  2. Dan Zhou2,
  3. Yimin Zhu1
  1. 1Department of Epidemiology & Biostatistics, School of Public Health, Zhejiang University, Hangzhou, China
  2. 2Department of pathophysiology, School of Medicine, Zhejiang University, Hangzhou, China
  1. Correspondence to Yimin Zhu, Department of Epidemiology & Biostatistics, School of Public Health, Zhejiang University, 866 Yuhangtang Road, Hangzhou, Zhejiang 310058, China; zhuym{at}


Background Metabolically healthy obese phenotype (MHO) refers to obese individuals with absence of metabolic abnormalities such as dyslipidaemia, insulin resistance and hypertension. Many studies reported the long-term prognosis of MHO on diseases and mortality with inconsistent results.

Methods We performed a meta-analysis to assess the risks of cardiovascular (CV) events and all-cause mortality for MHO individuals. Original prospective observational studies were searched in Medline, EMBASE, Web of Science and Cochrane library up to 30 September 2015. In this meta-analysis, the relative risk (RR) calculated on the basis of the incident number of disease events and deaths in participants and the corresponding multivariable-adjusted HR were both extracted to calculate pooled risk estimates. A random-effects model was used if there was heterogeneity among studies; otherwise, the fixed-effects model was used.

Results 22 prospective studies, involving 584 799 participants, were archived in the analyses. With metabolically healthy normal weight as the reference, the MHO phenotype was associated with a higher risk of CV events (RR 1.50, 95% CI 1.27 to 1.77; HR 1.60, 95% CI 1.38 to 1.84). However, MHO individuals were not associated with increased risk of all-cause mortality (RR 1.18, 95% CI 0.83 to 1.66; HR 1.07, 95% CI 0.92 to 1.25).

Conclusions The meta-analysis confirms a positive association between a metabolically healthy obese phenotype and the risk of CV events. However, higher risk for all-cause mortality is not evident in metabolically healthy obese individuals.

  • Epidemiology of cardiovascular disease
View Full Text

Statistics from


Recently, much interest has been focusing on the concept of ‘all obesity is not created equally’.1 Obese individuals without the metabolic abnormalities that commonly accompany excess adiposity (dyslipidemia, glucose dysregulation, and inflammation), termed metabolically healthy obesity (MHO), account for as much as 30–40% of the obese adult population.2 However, considerable controversy surrounds this issue, as the long-term outcomes of MHO on obesity-related diseases and mortality appear to be conflicting in relation to data from previous studies.3–8

In the past decade, a number of epidemiological studies had reported the risk of obesity-related diseases with the MHO phenotype; however, the strength of the association had varied with some studies reporting a twofold to threefold increase in the risk of type 2 diabetes, while other studies reported no risk or a slightly increased risk of cardiovascular (CV) disease among MHO individuals.9–14 A meta-analysis published in 2013 cast doubt on the existence of MHO.15 However, their results are controversial.16 ,17 The meta-analysis roughly merged the incidents of CV events and all-cause mortality together to calculate the pooled risk estimates for MHO individuals. However, it might not be reasonable. Although obesity was associated with increased risks of CV events, several studies found that increasing body mass index (BMI) had an opposite effect on all-cause mortality in elderly persons18 and patients with stroke19 and coronary artery disease.20 Another limitation of the meta-analysis was the fact that it did not adequately adjust for many potentially important baseline factors, including age, sex, physical activity and other critical elements.15

Owing to the dozens of prospective studies reporting the risk of CV events and mortality associated with the MHO phenotype with conflicting results and ongoing controversy on this issue, there is great need to systematically and comprehensively review evidence to define more accurate estimates of the relationship between the MHO phenotype and the prognosis. Therefore, we conduct a meta-analysis to estimate the effect of the MHO phenotype on the risk of CV events and all-cause mortality.

Materials and methods

Search strategy and selection criteria

This meta-analysis was conducted in accordance with the Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines.21 A systematic literature search was performed using the database of Medline, EMBASE, Web of Science and Cochrane library, and was supplemented through the manual review of the reference list of obtained articles up to Sep 30, 2015. The following truncated search terms were used: BMI, obesity, metabolic, metabolically, healthy, metabolic syndrome, CV disease, risk, mortality, cause of death.

All potentially eligible studies were considered for further review, which was carried out by one of the authors and then confirmed by another author. The two authors also retrieved and assessed potentially relevant publications, and the reference lists of the screened literatures as well as previous relevant reviews and meta-analyses were also checked to identify additional publications of interest.

Studies were considered eligible for meta-analysis if they met the following criteria: (1) the study had a prospective design, and the study population was adults (age ≥20) at baseline; (2) the weight classification of obesity and normal weight were defined on the basis of BMI and waist circumference cut-offs; (3) the study had stratified participants according to metabolic status and weight classification; (4) the outcomes of the study were CV events and all-cause mortality. The studies were excluded if they were literature reviews, cross-sectional studies, case–control studies, animal studies and genetic variation studies.

Data extraction and quality assessment

Two investigators independently carried out information extraction. The extracted data included publication data (first author's name, year of publication), country of the participants, the onset year of the study, the mean age of participants, definition of metabolic health, obese criteria, the length of follow-up, the number of participants who had an event according to BMI categories and metabolic status. If there was disagreement, the third investigator resolved it.

In this meta-analysis, study' quality was assessed by review of study design, including appropriate eligibility criteria of selection of participants, assessment of physical examination and measurement (self-report or clinical examination), assessment of outcomes (self-report or validated medical records), describing the characteristics of the participants, giving unadjusted estimates and confounder-adjusted estimates; making an overall interpretation of results considering objectives, limitations and results from similar studies. Each of the six quality criteria was evaluated and scored on an integer scale (0 or 1, with 1 being better) and summed. The scores below three were considered as low study quality. Quality assessment was performed independently by two authors, and discrepancies were solved by consensus or consultation with a third reviewer.

Definition of metabolic health and outcomes ascertainment

Most studies defined metabolically healthy status as the presence of less cardiometabolic abnormalities according to the metabolic syndrome diagnostic criteria of the Third Report of the National Cholesterol Education Program's Adult Treatment Panel (NCEP ATPIII) and International Diabetes Federation (IDF).22 ,23 The criteria of ATP III and IDF contained the same metabolically abnormal components, including elevated triglyceride, low high-density lipoprotein cholesterol, elevated blood pressure, or use of antihypertensive drug therapy, and elevated fasting plasma glucose or antidiabetic treatment. Two studies defined metabolic health on the basis of the homeostasis model assessment of insulin resistance (HOMA-IR) (insulin sensitive: HOMA-IR <2.5; insulin resistant: HOMA-IR ≥2.5).24 ,25 One study defined metabolic health involving a low level of inflammation.26

Most studies reported CV disease events as coronary heart disease events (new onset angina, fatal and non-fatal myocardial infarction, sudden cardiac death, fatal and nonfatal heart failure), stroke and CVD death. Two studies reported CV events containing heart failure or myocardial infarction.27 ,28 One prospective study which used the incident hypertension at any follow-up examination as end point events was also included in the meta-analysis, since the development of hypertension was a strong predictor of CVD mortality. 29

Data synthesis and analysis

The metabolically healthy normal-weight (MHNW) group was used as the reference group to assess the risks of CV events and all-cause mortality for MHO participants. In this meta-analysis, we used two strategies to calculate the effect estimates, the unadjusted relative risk (RR) calculated by the incident number of disease events and deaths and correspondingly multivariable-adjusted HR were extracted to calculate the overall risk estimates and 95% CIs.

The I2 statistic was used to describe the percentage of between-study heterogeneity. I2 described the percentage of total variation across studies that was due to heterogeneity rather than sampling error and ranges between 0% (no inconsistency) and 100% (high heterogeneity) with values of 25, 50 and 75% suggesting low, moderate and high heterogeneity.30 A random-effects model was used when heterogeneity was moderate or high; otherwise, the fixed-effects model was used. The sources of heterogeneity were detected by subgroup analysis and meta-regression using prespecified study-level characteristics as covariates, including follow-up time, participants' age at baseline, participants' age at end point, the Cox regression model in the original studies adjusted for physical activity or smoking, the allowed number of metabolic abnormality in metabolically healthy definition, the criteria used to define MHO, geographic location of the study, sample size. Sensitivity analysis excluding one study at a time was conducted to detect the influence of each study on the overall effect. Small study effects, such as publication bias, were assessed by inspecting the funnel plots for asymmetry and with Egger's test, with the results considered to indicate small study effects when p<0.10. All statistical analyses were carried out by STATA V.11.0 (STATA Corp, College Station, Texas, USA).


Characteristics of included studies

Our initial search identified 1587 potentially relevant articles, as shown in the flow chart (figure 1). After screening of the title/abstract and identifying 11 articles through manual search, 166 articles remained, of which 138 were excluded as the studies were not eligible. Following a detailed assessment, three studies were excluded for not using BMI value or waist circumference to identify the obese group and objective clinical measure to diagnose metabolic status.31 ,32 For two studies, we corresponded with authors, but data were not available.11 ,33 The quality assessment of remaining studies showed that two studies were excluded for low quality.7 ,34 Finally, 22 cohort studies met the inclusion criteria and were included, all of which were published in English.3–6 ,8–10 ,13 ,14 ,24–29 ,35–41 Eighteen studies evaluated CV events risk,4–6 ,8–10 ,13 ,14 ,24 ,26–29 ,35–38 ,41 and 11 studies evaluated all-cause mortality risk.3 ,4 ,8 ,13 ,24–26 ,35 ,38–40 The detailed information of included studies is presented in online supplementary file 1. The studies included in the meta-analysis represented a geographically diverse set of populations; however, ethnic composition was not specifically reported in most studies. Ten cohorts were constructed and followed up in Europe, seven in North America and five in Asia.

Figure 1

Outline of the study selection process for the meta-analysis.

Supplementary data 1

MHO phenotype and risk of CV events

Eighteen studies reported the association between the MHO phenotype and the risk of CV events and included 3903 cases and 178 687 participants (see online supplementary file 2). Pooled RR and HR were 1.50 (95% CI 1.27 to 1.77) and 1.60 (95% CI 1.38 to 1.84), respectively (figure 2). The heterogeneity among studies was significant in the pooled analysis with the unadjusted data set (I2=66.2%, p<0.001), but not significant with the adjusted data set (I2=45.6%, p<0.019).

Figure 2

Meta-analyses of metabolically healthy obese participants for the risk of CV events compared with metabolically healthy normal-weight individuals. RR, relative risk.

Supplementary data 2

Table 1 shows the results of the meta-regression and subgroup analyses for the prespecified covariates. No factors were found to account for the heterogeneity among subgroups. We observed that the MHO individuals were not at increased risk of future CV events with a follow-up duration of less than 10 years, whereas they showed a significantly higher risk with a longer follow-up duration (>10 years) (RR 1.58, 95% CI 1.39 to 1.81; HR 1.60, 95% CI 1.41 to 1.82). In the subgroup analysis, the studies were classified into three groups by the definition of metabolic health. Pooled analysis of the studies with the strictest criterion of metabolic health (meeting none of the cardiometabolic risk factors) showed that the MHO phenotype was not associated with the risk of CV events (RR 1.43, 95% CI 0.95 to 2.17; HR 1.17, 95% CI 0.28 to 4.91). However, the summary RR and HR showed positive association between MHO and CV events among the studies with less strict criteria (meeting less than two of the cardiometabolic risk factors). The pooled RRs and HRs did not vary within other subgroups, and the results were similar with the overall risk estimate.

Table 1

Cardiovascular events risk for metabolically healthy obesity in different subgroups compared with metabolically healthy normal weight

MHO phenotype and risk of all-cause mortality

Eleven studies assessed the risk of all-cause mortality associated with the MHO phenotype and included 2 705 deaths and 118 471 participants (see online supplementary table S2). The summary RRs were 1.18 (95% CI 0.83 to 1.66, I2=84.5%, p<0.001 for heterogeneity) with the unadjusted data set and 1.07 (95% CI 0.92 to 1.25, I2=17.6%, p=0.276 for heterogeneity) with the adjusted data set (figure 3).

Figure 3

Meta-analyses of metabolically healthy obese participants for the risk of all-cause mortality compared with metabolically healthy normal-weight individuals. RR, relative risk.

Table 2 presents the results of meta-regression and subgroup analysis. For the adjusted analysis, marginal statistical significance was observed when pooling the studies with a small sample size (<5000) (HR 1.40, 95% CI 1.05 to 1.87, I2=0.0%, p=0.880 for heterogeneity), but not significant for studies with a large sample size (>10 000) (HR 0.97, 95% CI 0.82 to 1.16, I2=32.4%, p=0.205 for heterogeneity). The pooled HR did not vary within other subgroups, and the results were similar with the overall risk estimate.

Table 2

All-cause mortality risk for MHO in different subgroups compared with metabolically healthy normal weight

Sensitivity analysis and publication bias

For sensitivity analysis, each single study was removed at a time and the analysis was repeated on the remaining studies to assess whether our findings were affected by the excluded studies. All the results appeared to be robust to the influence of the individual study (see online supplementary file 3).

No significant publication bias was detected by Egger's tests for all analysis; all p values for a two-sided test were >0.05 (see online supplementary file 4). There was also no apparent systematic bias as assessed by funnel plots for all comparisons performed (see online supplementary file 5).

Supplementary data 3

Supplementary data 4

Supplementary data 5


Although two meta-analysis studies had assessed the risks of CV diseases and mortality for the MHO phenotype.15 ,42 In their meta-analyses, they did not differentiate the incidents of CV events and all-cause death events in separate, but merged them together to calculate the pooled risk estimate. Whereas, we evaluated the risk of CV events and all-cause mortality for MHO individuals, respectively. Different with their conclusions, results of this study showed robust data that MHO was not associated with increased risk of all-cause mortality. Additionally, we reaffirmed that a metabolically healthy profile did not completely protect obese individuals from incident CV disease events.

On the basis of the present meta-analysis, it seemed that obesity was an independent predictor of CV events. One study reported that the MHO group had a significantly greater prevalence of coronary atherosclerosis compared with metabolically healthy normal-weight counterparts.43 It suggested that despite the metabolic parameters of the participants falling within the normally accepted range, it appeared that the positive association between obesity and subclinical coronary artery disease was still affirmed. Another study also found that MHO individuals had a significantly greater subclinical CVD burden, such as common carotid artery intima media thickness (IMT), aortic pulse wave velocity, and coronary and aortic calcification, compared with normal-weight groups.44 Since MHO was associated with double risk for T2D, it was also possible that some of the MHO individuals at baseline developed diabetes during follow-up, which in turn led to CV events.45 To support this, in a study with 20 years follow-up, approximately half of the healthy obese adults became unhealthy.46 Prospective investigation considering the changes of metabolically healthy status was needed to evaluate the risk of developing CV events for stable MHO individuals.

Several studies demonstrated that cardiorespiratory fitness (CRF) could play a central role in the prognosis of MHO individuals.3 ,47 In their prospective study, they observed that when regression models were not adjusted for CRF or physical activity, the results suggested that obesity per se (either MHO or MAO) was associated with a higher risk of CV events. However, the differences disappeared after adjustment for CRF or physical activity. In the present study, the same analytic strategy was performed. The result still presented statistical significance (HR 1.33, 95% CI 1.06 to 1.68) when combined the results of the original studies which had adjusted for physical activity in their regression model, however, it weakened the association between the MHO phenotype and the occurrence of CV events compared to the studies without accounting for physical activity. Several studies had consistently found that MHO individuals were associated with high levels of physical activity. Also, a higher level of CRF substantially reduced the adverse effects of obesity on morbidity.48 Previous studies found strong evidence indicating that better CRF levels were associated with fewer metabolic complications and lower risk of CV events at any age.49 It was also suggested that greater emphasis should be placed on improving CRF rather than weight loss per se in the primary and secondary prevention of CVD.50 However, almost half of the studies related to the prognosis of MHO individuals had not considered the role of physical activity or CRF, and a better understanding of the role of CRF in the relationship between MHO and outcomes might lead to better risk assessment.

Even though the MHO phenotype was associated with elevated risk in CV events, unexpectedly, it did not present significantly increased risk of all-cause mortality. Two meta-analyses assessed the association between BMI and all-cause mortality in elders without considering the presence of metabolic factors. Both of them did not find increased risk of mortality for obese participants.51 ,52 In the present meta-analysis, we found similar results. The risk of all-cause mortality was decreased for the MHO individuals in the age group of over 70 years old. It was possible that as age advanced, BMI decreased because of modifications of food intake and physical activity, hormone changes and weight loss.53 The MHO individuals were at an earlier stage of obesity-related disease progression, and the obese participants with advancing age yet still healthy at the time of data collection were most likely to lose weight and maintain that weight loss during the follow-up period, which would have maintained health status, thereby reducing risk of mortality. There was also another possibility. A number of studies had reported that in patients with CVD, the overweight and mildly obese were having a better prognosis than their leaner counterparts with the same CVD, a phenomenon termed as the ‘obesity paradox’.48 ,54 ,55 To support this, a large systematic review and meta-analysis of 97 studies reported that overweight and moderate obesity (BMI of 30–35 kg/m2) were associated with lower all-cause mortality even in the general population.56 The contribution of CRF was also important on all-cause mortality. Since MHO individuals had a higher level of CRF, it might attenuate the mortality risk associated with excess adiposity.48 Another meta-analysis also confirmed that individuals with preserved CRF have a good prognosis, regardless of BMI status.57 Several hypotheses have been proposed to explain the obesity paradox, including inability of BMI to differentiate between central and peripheral fat deposits, non-measurable confounding factors, possibility of selection bias and the effect of CRF.58 Prospective studies should combine the components of cardiometabolic risk factors and CRF to determine the impact of adiposity on morbidity and mortality.

Many investigators had described the particular characteristics that might distinguish MHO individuals from those at higher metabolic risk.49 However, there was still a lack of consensus among studies about how to define metabolic health. Most studies defined the metabolically healthy phenotype based on various combinations of four conventional metabolic criteria: blood pressure, high-density lipoprotein (HDL) cholesterol, triglycerides and fasting plasma glucose. Other components less commonly used to define MHO were HOMA-IR and C reactive protein.24 ,41 Most studies defined it as presenting less than one or two metabolic abnormalities. Few studies defined metabolic health as having none of the metabolic syndrome criteria aforementioned.29 ,59 The obese subjects possessed one or two cardiometabolic risk factors were classified as health at baseline, but they tended to develop more metabolic abnormalities during follow-up, and then leading to a CV event.60 We found that the meta-analysis with the studies which used the strictest definition (absence of all metabolic abnormality) obtained an insignificant association between MHO and CV events risk (RR 1.43, 95% CI 0.95 to 2.17; HR 1.17, 95% CI 0.28 to 4.91). However, the pooled analysis with the studies which used a less strict definition (relax to allow one or two metabolic abnormalities) proved a significant association. Several researchers questioned the current diagnosis of MHO, because chronical non-communicable disease was always a continuum rather than a categorical phenomenon. However, current studies processed it as separate healthy status at a fixed point, and it would ignore the variability of metabolic status during the follow-up. In the circumstance of disagreement of classification of metabolic health profile and imprecise measurement of obesity, it made the prediction of the prognosis of the MHO phenotype more difficult.

The strengths of this study included a large sample size, enabling the determination of robust estimates for the association between the MHO phenotype and related disease and mortality. Different with previous meta-analyses, we used both raw data and fully adjusted effect sizes from original studies to evaluate the risk of CV events and mortality for MHO individuals. The results were robust and reliable, for the similar and consistent results were obtained from the unadjusted and adjusted data sets. The limitations of this study should be considered. First, Wildman et al61 had shown that ethnic background was an important factor in determining MHO, and that non-Hispanic black participants had a higher prevalence of MHO. However, owing to lack of detailed information in these original studies, we only stratified the data by the geographic location of the study in the subgroup analysis. Weight gain was an important risk factor coupled with the development of metabolic abnormality. However, no studies reported the data regarding weight changes during the follow-up. It was possible that weight changes, rather than weight status at the time of data collection, impact mortality risk most profoundly. Finally, the publication bias might still be relevant even though best efforts were to conduct a comprehensive search and without statistical evidence of bias.

In conclusion, compared with metabolically healthy normal weight persons, MHO individuals are at increased risk for CV events, but they are not associated with a high risk of all-cause mortality. Nevertheless, the term ‘healthy’ may not be suitable to describe this population because of the possibility of underestimating the long-term effect of obesity, as well as less attention being paid to supporting the weight loss therapy. To reduce the risk of morbidity and mortality for MHO individuals, weight reduction and improvements in fitness are needed and beneficial.

What is already known on this subject

  • There was substantial evidence that metabolically healthy obese adults had increased risk of developing type 2 diabetes compared with metabolically healthy normal-weight adults.

  • Many studies have been conducted to assess the association of metabolically healthy obesity with cardiovascular disease events and all-cause mortality, and the results remain controversial.

What this study adds

  • This meta-analysis shows that metabolically healthy obesity is significantly associated with increased risk of cardiovascular disease events, but presents no increased risk of all-cause mortality.

  • Cardiorespiratory fitness seems to play an important role in the prognosis of metabolically healthy obesity individuals.


View Abstract


  • RZ and DZ contribute equally.

  • Contributors RZ and DZ conceived of the study, refined the study question and design, and extracted data. RZ drafted the manuscript. YZ refined the study design, obtained data, interpreted results, and contributed to the revision of the manuscript. All authors reviewed and edited the manuscript.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.